[Acpc-l] INFORMATIKKOLLOQUIUM: 26.11.2001 17h00
Gérard Le Lann: Is "Asynchronous Real-Time" An
Oxymoron (And Why Should You Care)?
Therese Schwarz
sek@dbai.tuwien.ac.at
Mon, 22 Oct 2001 11:12:08 +0200
Der Fachbereich Informatik und die Österreichische Computer Gesellschaft
laden zu folgendem Vortrag ein:
****************************************************************************
Is "Asynchronous Real-Time" An Oxymoron (And Why Should You Care)?
Gérard Le Lann
INRIA Rocquencourt, Frankreich
Zeit: 26.November 2001, 17:00 Uhr
Ort: Zemanek Hörsaal, Favoritenstraße 11/Erdgeschoß/roter Bereich
Abstract: When moving from the "capture" of some real world problem to the
specification of a matching problem in computer science, and then to the
specification of a correct design of a system-solution, four essential proof
obligations must be fulfilled.
Besides proofs related to safety, liveness, timeliness ("real-time"), and
dependability properties, others have to do with coverage, i.e. with the
accuracy of (1) models meant to reflect some future operational environment
and technology, (2) assumptions which underlie a design and its companion
proofs.
These proof obligations will be presented for the case of distributed
fault-tolerant
real-time computing problems, and special attention will be devoted to
computational models.
We will explore the following long-standing issue: "Does it make sense to
consider some asynchronous model when designing a solution to a real-time
computing problem?".
Concepts of "design immersion" and "late binding" (of a design to a model)
will
be presented. Shortcomings of synchronous or "timed" models will be explained,
as well as reasons why the intuitive belief that asynchrony and timeliness are
contradictory is mistaken.
We will demonstrate the following results:
- any real-time computing problem has asynchronous design solutions,
- regarding coverage, asynchronous design solutions dominate synchronous ones,
- asynchronous design solutions need not be less efficient than synchronous
ones,
- if safety properties must be maintained no matter what, then asynchronous
design
solutions are necessarily more efficient than synchronous ones.
If time permits, an example of how to design asynchronous real-time systems
will be given, as well as an illustration drawn from an on-going space project.
Biographie: Dr. Gerard Le Lann is Research Director at INRIA (Institut
National de
Recherche en Informatique et Automatique), Rocquencourt, France.
He holds French degrees, namely a M.S. in Applied Mathematics, an Engineer
Degree and a Ph.D, both in Computer Science.
G. Le Lann has been active in the field of computing since 1967, when he
completed his national duties with the French Navy, as a contributor to the
development of real-time multiprocessors intended for nuclear submarines.
Since 1969, he successively joined CERN (the European Research Center
for Nuclear Physics, Switzerland), the Cyclades Project (French Ministry
of Industry), inspired from the US Arpanet Project, and Stanford University,
where he worked with Professor Vint Cerf. There, he contributed to the design
of what became known as the TCP/IP protocol. In 1977, he published one of
the pioneering papers on distributed computing.
G. Le Lann has been with INRIA since 1978. There, he has launched three
projects so far, on distributed fault-tolerant real-time computing. Besides
conducting research work, he has been active in technology transfer. He is
regularly invited to present papers and to act as a consultant for various
international agencies and companies.
His current interests lie with theoretical issues in the area of asynchronous
real-time computing, and with more practical issues, such as proof-based
system engineering.
--
Schwarz Therese
Vienna University of Technology - Database & AI Group
A-1040 Vienna, Favoritenstr. 9-11/Stg.2/3. Stock/1842
http://www.dbai.tuwien.ac.at/staff/sek